Consistent misspellings

By Jack Stovell · published 2026-09-24

Consistent misspellings (consistent_misspellings) are the words a writer spells the same idiosyncratic way every time, on purpose or out of habit, such as an older form or a brand spelling. In a ScriptGrain voice profile the extraction model lists up to three per analysis, quoted verbatim from your samples. The catch: this attribute never moves your voice match score.

Analyse your consistent misspellings free

Your free voice profile measures consistent misspellings among all 45 attributes, from a sample of your own writing. No card needed.

Try the six-signal check first

At a glance

What consistent misspellings measures

This attribute catalogues the spellings you keep repeating that deviate from the dictionary norm. Not typos, not one-off slips: patterns. Maybe you write "greatful" every time instead of "grateful". Maybe your brand name uses a spelling nobody else would choose. The app's own definition is blunt about it: "idiosyncratic spellings that recur across samples". It sits in the Quirks and cadence group alongside four siblings, including rhythm pattern and structural signatures, the small tics that make a voice recognisably yours rather than generically correct.

The scale is simple: a list, capped at three per analysis, stored as a string array. The extraction model (Claude Sonnet 5) collects it in two passes, one per sample and one to synthesise the profile, and quotes each word verbatim, whatever its spelling. No scoring curve, no percentile.

How consistent misspellings is scored

It isn't scored at all. The voice-match engine has no comparator for this attribute, so it never moves a voice match score and polish can't target it either. Spelling idiosyncrasies are catalogued, not judged. ScriptGrain's reference corpus has no figure for them, so there is no typical value to quote, which suits something personal by definition.

So why keep it at all? Because the profile does more than score: ScriptGrain sends every attribute, this one included, to the model when it writes a draft. If you want the full picture of how attributes do or don't enter a score, the voice measurement framework walks through the mechanics; 28 of the 45 attributes can move a score, and this isn't one of them.

How to analyse consistent misspellings in your own writing

You can find your consistent misspellings by hand with a few samples and a dictionary.

  1. Gather three to five pieces of your own writing, ideally from different contexts (emails, posts, notes).
  2. Read through and flag any word spelled the same non-standard way more than once.
  3. Cross-check each flagged word against a dictionary for your chosen English variant, British or American.
  4. Discard autocorrect artefacts and any word that turns out to be standard in your variant.
  5. Keep only the words that repeat across samples: those are your consistent misspellings.
  6. Note whether each is deliberate (a brand name, an old form you prefer) or just a habit you'd never noticed.

ScriptGrain's free voice profile reads this straight from a writing sample, as one of all 45 voice attributes.

Examples of consistent misspellings in real writing

Example: a writer who always types "recieve" instead of "receive", in every sample, without fail.

Example: a brand that spells its own name "Colorwrks" and wants it kept that way everywhere.

Example: a columnist who writes the older "connexion" for "connection" in every piece, whatever the house style guide says.

None of these are errors in the writer's own frame. They're fingerprints. To be fair, some "consistent misspellings" turn out to be valid alternative spellings once you check which side of the Atlantic they're from; that's where a British vs American English markers reference earns its keep.

How to make AI keep your spellings

If you want an AI assistant to stop "correcting" your deliberate spellings, tell it plainly rather than hinting.

  1. List the specific spellings you use on purpose: older forms, house styles, brand names.
  2. Tell it which English you write in, British or American, so it isn't guessing.
  3. Ask it explicitly to leave those flagged words uncorrected, even where they look wrong.

When ScriptGrain writes, your profile goes into the system prompt under "Treat every attribute as a hard constraint, not a suggestion", so your listed spellings travel with it. The same profile reaches ChatGPT, Claude and other tools through the API and MCP server, so each of them can read the same list.

How to make AI use standard spelling

Sometimes you want the opposite: clean, standard spelling with no personal quirks creeping in.

  1. Set the English variant explicitly, British or American, rather than leaving it to default.
  2. Ask directly for consistent spelling throughout the piece.
  3. Run a spellcheck set to that variant once the draft is done.
  4. Keep proper nouns spelled as their owners spell them, regardless of variant.

When ScriptGrain writes, a variant block asks for British English (UK) "spelling, vocabulary, idioms, punctuation conventions, and date / number formats throughout", or American English, and says "Do NOT mix variants"; the variant you choose wins over the profile. The profile reaches ChatGPT, Claude and other tools through the API and MCP server.

Questions

What are consistent misspellings in writing?

They're spellings a writer repeats the same way across multiple samples, deviating from standard dictionary form. Could be a brand name, an old-fashioned variant, or a plain habit nobody's corrected. ScriptGrain's extraction model lists up to three per analysis, quoted verbatim from the writer's own samples, and files them in the Quirks and cadence group alongside attributes such as paragraph length.

What's idiosyncratic spelling in writing, and why keep it?

Idiosyncratic spelling is any recurring deviation from the norm that's specific to one writer rather than a general error pattern. Keeping it in a voice profile lets drafts keep it too: a brand spelling or a preferred older form is part of what makes writing sound like you, even though it has no comparator in the voice match score and polish can't target it.

Is there a British vs American spelling checker I can use?

Most word processors let you set the spellcheck language to UK or US English, which is the quickest check. When ScriptGrain writes, you set the variant explicitly (British or American) and its prompt tells the model "Do NOT mix variants". ScriptGrain's British vs American English markers reference explains the common divergences.

How do I make ChatGPT use British spelling?

Tell it directly: "use British English (UK) spelling, vocabulary and punctuation throughout, do not mix variants". Models often default to American spelling unless told otherwise, so make the instruction explicit and repeat it if the conversation runs long. ScriptGrain's own generation prompt sends a matching instruction whenever you choose British English.

How do I keep my own spelling quirks when writing with AI?

List the specific words you spell your own way, state which English variant you're using, and ask the tool to leave those words uncorrected. A vague request like "keep my style" gives the model nothing specific to protect. Naming the exact words does, and a ScriptGrain profile keeps up to three of them on file and sends them to the model whenever ScriptGrain writes a draft.

Related

Sources